Systems and methods for processing real-time video from a medical image device and detecting objects in the video

A computer-implemented system for detecting a feature-of-interest in a video, the system comprising:

• a discriminator network that includes a perception branch and an adversarial branch, the perception branch being configured to output detections of the feature-of-interest in the video;
• and
• a generative network configured to receive detections of the feature-of-interest from the perception branch of the discriminator network and generate artificial representations of the feature-of-interest based on the detections from the perception branch, the artificial representations including false representations that appear as true representations of the feature-of-interest,
• wherein the adversarial branch is configured to be trained by applying the artificial representations of the feature-of-interest to the adversarial branch and to provide an output identifying differences between the false representations and true representations of the feature-of-interest, and wherein the perception branch is further configured to be trained by the output of the adversarial branch so that false representations are not detected by the perception branch as true representations.

A computer-implemented method for detecting a feature-of-interest in a video, the method comprising the following steps performed by at least one processor:

• outputting detections of the feature-of interest by applying the video to a trained perception branch of a discriminator network;
• generating artificial representations of the feature-of interest by applying the detections of the perception branch to a generative network, the artificial representations including false representations that appear as true representations of the feature-of-interest;
• training an adversarial branch of the discriminator network by applying the artificial representations of the feature-of-interest to the adversarial branch, wherein the adversarial branch is configured to provide an output identifying differences between the false representations and true representations of the feature-of-interest;
• and
• further training the perception branch by applying the output of the adversarial branch to the perception branch so that false representations are not detected by the perception branch as true representations.